HTRA1/lncRNA HTRA1-AS1 dominates in age-related macular degeneration reticular pseudodrusen genetic risk with no complement involvement

Abstract

Age-related macular degeneration (AMD) is a multifactorial retinal disease with a large genetic risk contribution. Reticular pseudodrusen (RPD) is a sub-phenotype of AMD with a high risk of progression to late vision threatening AMD. In a genome-wide association study of 2165 AMD+/RPD+ and 4181 AMD+/RPD- compared to 7639 control participants, both chromosomes 1 (CFH) and 10 (ARMS2/HTRA1) major AMD risk loci are reidentified. However association is only detected for the chromosome 10 locus when comparing AMD+/RPD+ to AMD+/RPD- cases. The chromosome 1 locus is notably absent. The chromosome 10 RPD risk region contains a long non-coding RNA HTRA1-AS1 (ENSG00000285955/BX842242.1) which colocalizes with genetic markers of retinal thickness. HTRA1-AS1 has a strong retinal eQTL signal, pinpointing the parafoveal photoreceptor outer segment layer. Whole genome sequencing of phenotypically extreme RPD cases identifies even stronger enrichment for the chromosome 10 risk genotype.

Fig. 1. Overall AMD+/RPD+ compared to AMD+/RPD- study design in multiple stepwise strategies from cohort collection, GWAS, to subsequent post-GWAS analyses.

Age-related macular degeneration AMD, Reticular pseudodrusen RPD, Single nucleotide polymorphism SNP, Quality control QC, Genome-wide associations study GWAS, expression quantitative trait loci eQTL. Created in Biorender. Farashi, S. (2025) https://biorender.com/p05m3ap with copyright under a CC BY licence.

Fig. 2. Results of Genome-wide association studies (GWAS).

a Manhattan plot of overall age-related macular degeneration (AMD) GWAS (baseline) showing –log10(P-value) values across the genome comparing AMD+ (reticular pseudodrusen (RPD)+/-) participants (N = 3276) vs. AMD- controls (N = 1946). The red horizontal line represents the genome-wide significance level (P-value  =  5 × 10⁻⁸). Nearby genes to the two main associated loci with AMD risk on chromosomes 1, 6 and 10 are labelled; b Manhattan plot of the meta-analysis of AMD+/RPD+ participants compared to AMD+/RPD- participants (GWAS1). Common variants display a genome-wide significant locus on chromosome 10. The y-axis indicates −log10(P-value). The red and blue horizontal lines represent the genome-wide significance level (P-value  =  5 × 10⁻⁸) and suggestive genome-wide significance level (P-value  =  1 × 10⁻⁶), respectively. Note that the RPD-specific GWAS found only the ARMS2/HTRA1 risk locus on chromosome 10, where the complement cascade risk factors on chromosome 1 (CFH/CFHRs) were absent; c Regional plot with annotations displaying the top RPD-associated signal. The color of each point (single nucleotide polymorphism; SNP) is determined by its linkage disequilibrium (LD) squared correlation coefficient (r²) with respect to the top lead variant, displayed in dark purple (rs11200638). Independent significant SNPs are displayed as red circles outlined in black. Genes proximal to this RPD risk locus are labeled, including ARMS2, HTRA1 (both in red) and the long non-coding RNA HTRA1-AS1 (in black). HTRA1-AS1 was added manually. The orange dashes over the top of the SNPs indicate SNPs in the reference genome (1000 G Phase 3 European ancestry). Genomic locations are plotted using the hg19 genomic build. Generated using the FUMA GWAS web tool (v1.5.3) (https://fuma.ctglab.nl) created by Watanabe et al., Functional mapping and annotation of genetic associations with FUMA. Nat. Commun. 8:1826. (2017). Note: only individual-level participants (AMD+/RPD(+/-) N = 3276 versus AMD-/RPD- N = 1946; Total N = 5222) were included in Fig. 2a (as data from summary statistics cohorts was not available), whereas in Fig. 2b both individual-level participants and participants with summary statistics were included (AMD+/RPD+ N = 2165 versus AMD+/RPD- N = 4181; Total N = 6346). See Supplementary Fig. 9 for RPD-specific GWAS analysis of only individual-level participants, which shows the same finding as the grouped data in (b). All P-values are uncorrected, based on a two-sided Wald test.

Fig. 3. Reticular pseudodrusen (RPD) extent is significantly associated with heightened RPD genetic risk.

a Box plot showing the RPD extent in N = 526 individuals relative to the RPD-risk lead single nucleotide polymorphism (SNP)(rs11200638) dosage (x-axis). b Box plot showing RPD extent in N = 526 individuals relative to the CFH top risk SNP (rs10922109) based on Fritsche et al. dosage (x-axis) while adjusting for age and sex as covariates. In a and b, the y-axis represents the log-transformed RPD extent based on the proportion of the entire volume scan with RPD (the mean RPD extent of the left and right eyes was used) adding a constant of 0.1 to avoid log(0) of N = 181 individuals who showed a mean extent RPD value of zero. A linear regression test was conducted to examine the differences in the RPD extent across the genotype groups controlling for age and sex. We performed pairwise two-sided Wilcoxon tests with Bonferroni correction to directly compare the P-values among groups. Significant differences between all pairs of groups are indicated by asterisks (P-value > 0.05: not significant (ns), P-value = 0.046: *, P-value = 0.0007: ***); c RPD extent as defined in panel a (y-axis) for N = 526 individuals, binned into deciles (x-axis) for analysis of RPD risk. d RPD risk allele frequency (left y-axis) relative to RPD extent deciles (x-axis). In d, blue points represent individual study participants’ genetic risk, colored by risk allele dosage. Orange points represent risk allele frequency per decile (right y-axis). The plot illustrates how the RPD-risk lead SNP genotypes vary across different levels of RPD extent, providing insights into the genetic association between RPD risk and the severity of RPD. In boxplots, a–c each box shows the median (middle line), bounds of boxes show the interquartile range (25th–75th percentiles) and whiskers extend to the most extreme data-points within ±1.5× interquartile range from respective quartiles; outliers are shown as points. There are no replicates in this dataset.

Fig. 4. Increased reticular pseudodrusen (RPD) genetic risk is associated with increased expression of HTRA1-AS1 retinal RNA, and decreased expression of HTRA1 in multiple tissues and independent datasets.

HTRA1-AS1: long non-coding RNA Antisense 1 gene; HTRA1: HtrA serine peptidase 1. Datapoints represent gene expression and risk genotype dosage measured in post-mortem donor retinae. The y-axis represents the log-transformed expression level of each gene; the x-axis denotes the risk allele dosage of the RPD risk lead variant (0: individuals who carry no RPD risk allele, 1: heterozygous for RPD risk allele and 2: homozygous for RPD risk allele). Colour denotes age-related macular degeneration (AMD; teal) or non-AMD/control (pink). Regression testing is corrected for disease status (single regression line). Boxplots display the following information: middle line = median, upper box bound = 75th percentile, lower box bound = 25th percentile; whiskers extend to the most extreme data-points within ±1.5 × interquartile range from respective quartiles; outliers are included as points. NEI cohort: N = 310 AMD and N = 94 non-AMD donors; Genentech cohort: N = 23 AMD and N = 98 non-AMD donors. Asterisks at top right of each plot indicate a significant eQTL result (two-sided Student’s T test; Nominal P-value < 0.05 for NEI cohort; Corrected P-value < 0.05 for Genentech cohort, as eQTLs were tested in four retinal tissues). Full summary statistics including exact P-values are provided in Supplementary Tables 13 and 14.

Fig. 5. Genome-wide multi-trait colocalization analysis of reticular pseudodrusen (RPD) and candidate traits that show ARMS2/HTRA1 locus as a significant risk-associated region.

Stacked association plots of traits with >95% shared variants with RPD. Colocalization analysis with Coloc implicates the RPD risk lead variant rs11200638 as being the shared causal variant between RPD risk and expression quantitative trait loci (eQTLs), photoreceptor cell outer segment (OS) thickness and total retinal thickness. In contrast, the thickness of photoreceptor cell layers including the inner segment (IS) and photoreceptor cell outer nuclear layer (ONL), did not show a high overlap with the RPD risk region. Each plot includes a recombination (recomb.) rate track (blue line) and a linkage disequilibrium (LD) color scale indicating the r² values of single nucleotide polymorphisms (SNPs) in relation to RPD-risk lead SNP (rs11200638). The x-axis represents the position on chromosome 10, and the y-axis represents the -log10(P-value) of association. The blue bars on the x-axis label the genes. lncRNA HTRA1-AS1 was added manually. The purple diamond represents the most significant SNP in the region for each trait and the dark blue square represents the RPD-risk lead SNP (rs11200638) in each trait. See Supplementary Table 15 for P1, P2 and P12 values of sensitivity analyses. P-values for all traits except the RPD GWAS are based on two-sided Student’s T tests. P-values for the RPD GWAS are based on a two-sided Wald test. Note: The gene track represents annotated genes in the region, regardless of whether they are target genes for eQTLs.